# BCI-001: Brain-computer interfaces for communication and control > Generated from a local paper card. Do not edit this page directly; edit the source card and rerun `vp run docs:generate`. > Internal wiki boundary: paper-card evidence and user-provided PDF downloads are available for private reading. Extracted full-paper text is not published. > Source: `library/paper_cards/BCI-001.md` ## Paper Access * Internal PDF: Download Paper * DOI / official page: [10.1016/S1388-2457(02)00057-3](https://doi.org/10.1016/S1388-2457\(02\)00057-3) * Deployment boundary: these PDF links are intended only for a private/protected internal wiki. ## BCI-001: Brain-computer interfaces for communication and control ## Metadata * ID: BCI-001 * Title: Brain-computer interfaces for communication and control * Year: 2002 * DOI / URL: 10.1016/S1388-2457(02)00057-3 * Local PDF: see Paper Access section above * Text artifact: local-only path withheld from docs site * Review status: `extracted` ## Study Type * Track: BCI / EEG Foundations * Task: invited review of BCI communication and control systems * Participants or dataset: review paper; no single participant cohort or dataset * Hardware: scalp EEG, epidural/subdural recording, and intracortical recording are reviewed * Channels or sensors: EEG and related electrophysiological recordings; exact channel montage varies by reviewed BCI type ## Methods * Paradigm: BCI systems translating electrophysiological features into messages or device commands * Signal processing or model: feature extraction from evoked potentials, SCPs, mu/beta rhythms, or neuronal firing; translation algorithms map features to device commands * Training/calibration: user and BCI adapt to each other initially and continuously; feedback is central to operation * Online/offline: emphasizes online BCI operation and warns that offline analyses must be validated online ## Results * Metrics: review reports current BCI maximum information transfer rates up to about 10-25 bits/min; also discusses speed, accuracy, bit rate, and user/system performance as evaluation metrics * Main findings: BCI is framed as a non-muscular communication/control channel; successful operation depends on matching user intent, signal features, translation algorithms, feedback, and application demands * Reported limitations: low information-transfer capacity, artifact risks, need for long-term assessment, user variability, and need to match applications to users ## Relevance To This Project * Supports: the project framing that non-invasive EEG should be treated as a limited-capacity intent/control channel rather than a high-bandwidth low-level robot controller * Conflicts with: no direct conflict; the paper does not evaluate scene-aware BCI, YOLO-generated targets, or robotic grasping * Design implication: use EEG for high-level choices, confirmations, mode changes, and safety gates; evaluate both user signal control and system task performance ## Extracted Evidence | Claim | Status | Evidence Note | Page/Section | | --- | --- | --- | --- | | A BCI is a non-muscular channel for sending messages or commands to the external world. | verified | The abstract and definition sections describe BCI as translating electrophysiological activity into communication or device commands without conventional muscular output. | Abstract; Section 2.1 | | BCI operation is an adaptive skill rather than passive mind reading. | verified | The paper states that effective BCI operation depends on feedback and on interaction between two adaptive controllers: the user's brain and the BCI translation system. | Section 2.2; Fig. 1 caption | | Current BCI capacity supports basic communication/control but is limited for complex continuous control. | verified | The review reports maximum information transfer rates around 10-25 bits/min and says more complex applications require greater speed and accuracy. | Abstract; Section 5 | | The proposed system should evaluate both user-level signal control and system-level task performance. | verified | The paper distinguishes user performance from system performance and recommends speed, accuracy, information transfer rate, and real-life testing. | Sections 4.9.3-4.9.4 | | EEG-based BCI experiments must control for non-CNS artifacts and signal-to-noise limits. | verified | The review highlights EMG/EOG artifacts and signal-to-noise ratio as central issues for feature extraction and credible BCI control. | Section 4.5; Section 5 | | For SAH-BRI-Grasp, EEG should be scoped to high-level intent, confirmation, and supervisory commands rather than dense low-level robot motion. | inferred | This is an engineering inference from the paper's limited BCI bit-rate evidence, adaptive-skill framing, and application-matching guidance. | Abstract; Sections 4.9.3-5 | ## Open Questions * The review is from 2002; newer BCI performance limits must be checked against later BCI and robotic-arm papers. * It does not directly address SSVEP over dynamic object boxes or shared-control grasping. * The high-level-intent design implication should be triangulated with `MI-005`, `BRI-002`, `SA-001`, and `SA-002`.